I am using Netpyne (which is extention of NEURON). It has model specification file in python and also some of its Plots etc
Also I am using EPFL L23PC neuron model and point neurons as a source of IB input, OSC clock input and Interneuron GABA reset after OSC clock input
Current full model:
With right parameters this model works for 280+ signals out of 340. While some errors are by chance, some are biased errors and stay with different parameters and synapse types. Such errors are extra firing after pause and depressing after continuous firing.
To reduce error rate to ideally 0% I am trying to break sequence of signals into an array of more simple & typical L23PC neuron responces separated by OSC clock signals. This is done by GABA reset input
After each clock osc signal puts L23PC neuron into 2 states [YES or NO]. Ideally that states will be the same for either YES or NO
YES-STATE: After IB signal to depolarize L23PC to the same level (f.e. -30mv). Leave in this state (with slow decay) until OSC clock signal. After clock signal increase additional 30mv and fire AP
NO-STATE: (No IB) flat line at -65mv until OSC clock signal. After clock signal increase additional 30mv to -35mv. No AP
To achieve these states:
1 After each OSC signal GABA interneuron resets everything back to the flat line. all previous history by firing GABA interneuron reset
2 Memory about most recent IB signal is maintained via slow decay NMDA
Most problems come from synapse on basal dendrite AMPANMDA_EMS or exc all give some extra firing and attenuation with time. (RESULTS DIFFER DEPENDING ON SYNAPSE AND PARAMS) This part creates errors in 50-60 signals from 340. in this illustration - reproduce this error.
only basal dendrite receives signal from IB (green) and only GABA osc clock signal (red)
Here the scheme is only
this Illustration1 is just a way to represent 12MB output data file. Parameters from basal dendrite plotted
App4
On the right are params of AMPANMDA_EMS synapse. On the left are V potentials from dend9>dend7>dend0>soma>apic0>apic6. It illustrates how EPSPs travel along neuron processes. This is done to make sure that V wave comes from basal dendrites and not from some other parts such as soma, apical dendrites etc (in Illustration2.html EPSPs travel from apic and soma as well)
Causes of signal error:
1 at time=820. There is no change in params that influence synapse. Also there are no EPSPs coming from soma or apic. but there is change in v,g . (Which means parameters and other values are recalculated somewhere in NEURON core every time)
2 every consequent EPSP attenuates. probably because of depressing synapse or L23PC. Changing AMPANMDA_EMS synapse with default 'exc' PARTIALLY IMPROVED SIGNAL QUALITY. Probably some other synapse parameters would make EPSPs mor constant and tonic
3 strange values for gNMDA and iNMDA parameters. They are scaled down to e-200 & e-198. I/g gives -65v. However e-200 seem to be too unreal for EPSC
Illustration2.html shows the full model output (apic dendrites OSC clock input included). However apic input seems to give less signal error and its error mostly reflects basal error
ProbAMPANMDA_EMS
COMMENT
/*
Copyright (c) 2015 EPFL-BBP, All rights reserved.
THIS SOFTWARE IS PROVIDED BY THE BLUE BRAIN PROJECT ``AS IS''
AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO,
THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE BLUE BRAIN PROJECT
BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR
BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY,
WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE
OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN
IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
This work is licensed under a
Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International License.
To view a copy of this license, visit
http://creativecommons.org/licenses/by-nc-sa/4.0/legalcode or send a letter to
Creative Commons,
171 Second Street, Suite 300,
San Francisco, California, 94105, USA.
*/
ENDCOMMENT
TITLE Probabilistic AMPA and NMDA receptor with presynaptic short-term plasticity
COMMENT
AMPA and NMDA receptor conductance using a dual-exponential profile
presynaptic short-term plasticity as in Fuhrmann et al. 2002
_EMS (Eilif Michael Srikanth)
Modification of ProbAMPANMDA: 2-State model by Eilif Muller, Michael Reimann, Srikanth Ramaswamy, Blue Brain Project, August 2011
This new model was motivated by the following constraints:
1) No consumption on failure.
2) No release just after release until recovery.
3) Same ensemble averaged trace as deterministic/canonical Tsodyks-Markram
using same parameters determined from experiment.
4) Same quantal size as present production probabilistic model.
To satisfy these constaints, the synapse is implemented as a
uni-vesicular (generalization to multi-vesicular should be
straight-forward) 2-state Markov process. The states are
{1=recovered, 0=unrecovered}.
For a pre-synaptic spike or external spontaneous release trigger
event, the synapse will only release if it is in the recovered state,
and with probability u (which follows facilitation dynamics). If it
releases, it will transition to the unrecovered state. Recovery is as
a Poisson process with rate 1/Dep.
This model satisfies all of (1)-(4).
ENDCOMMENT
COMMENT
/**
@file ProbAMPANMDA_EMS.mod
@brief Probabilistic AMPA and NMDA receptor with presynaptic short-term plasticity
@author Eilif Muller, Michael Reimann, Srikanth Ramaswamy, James King @ BBP
@date 2011
*/
ENDCOMMENT
NEURON {
THREADSAFE
POINT_PROCESS ProbAMPANMDA_EMS
RANGE tau_r_AMPA, tau_d_AMPA, tau_r_NMDA, tau_d_NMDA
RANGE Use, u, Dep, Fac, u0, mg, Rstate, tsyn_fac, u
RANGE i, i_AMPA, i_NMDA, g_AMPA, g_NMDA, g, e, NMDA_ratio
RANGE A_AMPA_step, B_AMPA_step, A_NMDA_step, B_NMDA_step
NONSPECIFIC_CURRENT i
POINTER rng
RANGE synapseID, verboseLevel
}
PARAMETER {
tau_r_AMPA = 0.2 (ms) : dual-exponential conductance profile
tau_d_AMPA = 1.7 (ms) : IMPORTANT: tau_r < tau_d
tau_r_NMDA = 0.29 (ms) : dual-exponential conductance profile
tau_d_NMDA = 43 (ms) : IMPORTANT: tau_r < tau_d
Use = 1.0 (1) : Utilization of synaptic efficacy (just initial values! Use, Dep and Fac are overwritten by BlueBuilder assigned values)
Dep = 100 (ms) : relaxation time constant from depression
Fac = 10 (ms) : relaxation time constant from facilitation
e = 0 (mV) : AMPA and NMDA reversal potential
mg = 1 (mM) : initial concentration of mg2+
mggate
gmax = .001 (uS) : weight conversion factor (from nS to uS)
u0 = 0 :initial value of u, which is the running value of release probability
synapseID = 0
verboseLevel = 0
NMDA_ratio = 0.71 (1) : The ratio of NMDA to AMPA
}
COMMENT
The Verbatim block is needed to generate random nos. from a uniform distribution between 0 and 1
for comparison with Pr to decide whether to activate the synapse or not
ENDCOMMENT
v (mV)
i (nA)
i_AMPA (nA)
i_NMDA (nA)
g_AMPA (uS)
g_NMDA (uS)
g (uS)
factor_AMPA
factor_NMDA
A_AMPA_step
B_AMPA_step
A_NMDA_step
B_NMDA_step
rng
: Recording these three, you can observe full state of model
: tsyn_fac gives you presynaptic times, Rstate gives you
: state transitions,
: u gives you the "release probability" at transitions
: (attention: u is event based based, so only valid at incoming events)
Rstate (1) : recovered state {0=unrecovered, 1=recovered}
tsyn_fac (ms) : the time of the last spike
u (1) : running release probability
}
STATE {
A_AMPA : AMPA state variable to construct the dual-exponential profile - decays with conductance tau_r_AMPA
B_AMPA : AMPA state variable to construct the dual-exponential profile - decays with conductance tau_d_AMPA
A_NMDA : NMDA state variable to construct the dual-exponential profile - decays with conductance tau_r_NMDA
B_NMDA : NMDA state variable to construct the dual-exponential profile - decays with conductance tau_d_NMDA
}
INITIAL{
LOCAL tp_AMPA, tp_NMDA
Rstate=1
tsyn_fac=0
u=u0
A_AMPA = 0
B_AMPA = 0
A_NMDA = 0
B_NMDA = 0
tp_AMPA = (tau_r_AMPA*tau_d_AMPA)/(tau_d_AMPA-tau_r_AMPA)*log(tau_d_AMPA/tau_r_AMPA) :time to peak of the conductance
tp_NMDA = (tau_r_NMDA*tau_d_NMDA)/(tau_d_NMDA-tau_r_NMDA)*log(tau_d_NMDA/tau_r_NMDA) :time to peak of the conductance
factor_AMPA = -exp(-tp_AMPA/tau_r_AMPA)+exp(-tp_AMPA/tau_d_AMPA) :AMPA Normalization factor - so that when t = tp_AMPA, gsyn = gpeak
factor_AMPA = 1/factor_AMPA
factor_NMDA = -exp(-tp_NMDA/tau_r_NMDA)+exp(-tp_NMDA/tau_d_NMDA) :NMDA Normalization factor - so that when t = tp_NMDA, gsyn = gpeak
factor_NMDA = 1/factor_NMDA
SOLVE state
mggate = 1 / (1 + exp(0.062 (/mV) * -(v)) * (mg / 3.57 (mM))) :mggate kinetics - Jahr & Stevens 1990
g_AMPA = gmax*(B_AMPA-A_AMPA) :compute time varying conductance as the difference of state variables B_AMPA and A_AMPA
g_NMDA = gmax*(B_NMDA-A_NMDA) * mggate :compute time varying conductance as the difference of state variables B_NMDA and A_NMDA and mggate kinetics
g = g_AMPA + g_NMDA
i_AMPA = g_AMPA*(v-e) :compute the AMPA driving force based on the time varying conductance, membrane potential, and AMPA reversal
i_NMDA = g_NMDA*(v-e) :compute the NMDA driving force based on the time varying conductance, membrane potential, and NMDA reversal
i = i_AMPA + i_NMDA
}
NET_RECEIVE (weight,weight_AMPA, weight_NMDA, Psurv, tsyn (ms)){
LOCAL result
weight_AMPA = weight
weight_NMDA = weight * NMDA_ratio
: Locals:
: Psurv - survival probability of unrecovered state
: tsyn - time since last surival evaluation.
INITIAL{
tsyn=t
}
: Do not perform any calculations if the synapse (netcon) is deactivated. This avoids drawing from the random stream
if( !(weight > 0) ) {
VERBATIM
return;
ENDVERBATIM
}
: calc u at event-
if (Fac > 0) {
u = u*exp(-(t - tsyn_fac)/Fac) :update facilitation variable if Fac>0 Eq. 2 in Fuhrmann et al.
} else {
u = Use
}
if(Fac > 0){
u = u + Use*(1-u) :update facilitation variable if Fac>0 Eq. 2 in Fuhrmann et al.
}
: tsyn_fac knows about all spikes, not only those that released
: i.e. each spike can increase the u, regardless of recovered state.
tsyn_fac = t
: recovery
if (Rstate == 0) {
: probability of survival of unrecovered state based on Poisson recovery with rate 1/tau
Psurv = exp(-(t-tsyn)/Dep)
result = urand()
if (result>Psurv) {
Rstate = 1 : recover
if( verboseLevel > 0 ) {
printf( "Recovered! %f at time %g: Psurv = %g, urand=%g\n", synapseID, t, Psurv, result )
}
}
else {
: survival must now be from this interval
tsyn = t
if( verboseLevel > 0 ) {
printf( "Failed to recover! %f at time %g: Psurv = %g, urand=%g\n", synapseID, t, Psurv, result )
}
}
}
if (Rstate == 1) {
result = urand()
if (result
: release!
tsyn = t
Rstate = 0
A_AMPA = A_AMPA + weight_AMPA*factor_AMPA
B_AMPA = B_AMPA + weight_AMPA*factor_AMPA
A_NMDA = A_NMDA + weight_NMDA*factor_NMDA
B_NMDA = B_NMDA + weight_NMDA*factor_NMDA
}
else {
if( verboseLevel > 0 ) {
printf("Failure! %f at time %g: urand = %g\n", synapseID, t, result )
}
}
}
}
PROCEDURE setRNG() {
VERBATIM
{
/**
* This function takes a NEURON Random object declared in hoc and makes it usable by this mod file.
* Note that this method is taken from Brett paper as used by netstim.hoc and netstim.mod
* which points out that the Random must be in uniform(1) mode
*/
void** pv = (void**)(&_p_rng);
if( ifarg(1)) {
*pv = nrn_random_arg(1);
} else {
*pv = (void*)0;
}
}
ENDVERBATIM
}
FUNCTION urand() {
VERBATIM
double value;
if (_p_rng) {
/*
:Supports separate independent but reproducible streams for
: each instance. However, the corresponding hoc Random
: distribution MUST be set to Random.negexp(1)
*/
value = nrn_random_pick(_p_rng);
//printf("random stream for this simulation = %lf\n",value);
return value;
}else{
ENDVERBATIM
: the old standby. Cannot use if reproducible parallel sim
: independent of nhost or which host this instance is on
: is desired, since each instance on this cpu draws from
: the same stream
value = scop_random(1)
VERBATIM
}
ENDVERBATIM
urand = value
}
FUNCTION toggleVerbose() {
verboseLevel = 1-verboseLevel
}
